Moderation & Flagging Protocols

UpdatedSeptember 9, 2026

ByEpsteinWiki

How EpsteinWiki receives warnings, protects people, preserves evidence, reviews disputed content, and records decisions without turning moderation into censorship or accusation

A single bad sentence can do several kinds of harm at once. It can expose a survivor, misidentify an uninvolved person, distort a court record, amplify a threat, or make a genuine document look less trustworthy than it is. Moderation is where EpsteinWiki slows that chain reaction down.

This protocol governs what happens after a reader, contributor, editor, subject, survivor, attorney, or member of the public raises a concern. It is designed for a difficult archive: one built from litigation, testimony, government disclosures, leaked or disputed material, financial records, media, and records in which innocent contact can appear beside serious allegations.

The goal is not to erase uncomfortable facts. The goal is to make sure that what remains public is accurate, proportionate, lawfully handled, clearly attributed, and safe enough to publish.

Snapshot

A flag is a request for review, not a finding that content is false, unlawful, or abusive.
Anyone may flag a page, file, comment, account, edit, image, or pattern of conduct.
Immediate danger, credible threats, exposed survivor or minor identities, suspected child sexual abuse material, and active doxxing receive the highest priority.
Moderators may temporarily restrict content while reviewing it. A temporary restriction is not a final judgment.
Reviewers preserve lawful evidence, minimize further exposure, identify the exact claim at issue, and compare it with the strongest available sources.
Living people must not be described as criminals or participants in abuse without sources that support the exact statement.
A name in a contact book, calendar, email, photograph, financial record, or flight log proves only what that record proves.
Contributors must be told what rule or evidence issue affected their work, unless safety, privacy, or law prevents disclosure.
Material decisions require a written record. High-risk decisions should receive a second review.
Retaliation against a flagger, contributor, survivor, source, editor, or moderator is prohibited.

Purpose

EpsteinWiki exists to document a historical record that has too often been obscured by secrecy, institutional failure, sensationalism, and misinformation. That mission requires publication, but it also requires restraint.

This protocol gives moderators a repeatable way to answer five questions:

Is anyone in immediate danger?
Is the material legal and safe to possess or display?
Does the article accurately state what the evidence proves?
Is publication proportionate to the public interest?
What action creates the smallest necessary restriction while correcting the actual problem?

The protocol supports the Editorial Standards, Corrections Policy, Contributor Instructions, and Volunteer Conduct Code. When rules overlap, the more protective safety or privacy rule controls until a senior editor completes review.

Scope

These rules apply to:

Published articles and drafts
Uploaded documents, images, audio, and video
Titles, captions, excerpts, summaries, tags, and metadata
Comments, talk pages, submission forms, and direct messages sent through EpsteinWiki systems
Contributor accounts and moderator actions
Links to third-party material
Bulk imports, OCR text, data extracts, and automated summaries
Social posts published in EpsteinWiki’s name
Off-site conduct when it presents a credible safety threat, impersonates EpsteinWiki, compromises site systems, or targets someone because of their participation here

Editorial disagreement alone is not misconduct. A contributor may challenge a source, interpretation, or policy in good faith. Moderation begins when a concern involves accuracy, attribution, privacy, legality, safety, manipulation, harassment, system integrity, or repeated refusal to follow established standards.

Governing Principles

Every moderation decision should reflect these principles:

Safety first. Protect people before debating prose when there is a credible risk of harm.

Evidence before inference. Review the underlying record, not merely the confidence of the writer or complainant.

Minimum necessary intervention. Correct a sentence when a sentence is wrong. Restrict a file when the file is unsafe. Do not erase an entire article because one claim needs repair.

Trauma-informed handling. Do not force survivors to repeat unnecessary details, confront alleged perpetrators, or prove distress.

Due process. Give affected contributors a meaningful explanation and opportunity to respond when doing so is safe and appropriate.

Independence. No person should control a review involving their own work, conflict, friend, adversary, or outside interest.

Reversibility. Temporary measures should be easy to reverse if review clears the content.

Documentation. A decision that cannot be reconstructed later is not a complete moderation decision.

What a Flag Means

A flag means: Please examine this specific content or conduct under the published rules.

It does not mean:

The allegation in the flag is true
The contributor acted maliciously
The subject of an article is innocent or guilty
The content must automatically be removed
The flagger owns the editorial decision
A legal demand is valid merely because it uses legal language

Moderators classify the report, preserve what is necessary, reduce immediate risk, and investigate. The final result may be no action, correction, context, redaction, restriction, removal, account action, or referral outside the site.

Who May Flag Content

Anyone may submit a flag. No account, legal representative, or personal relationship to the subject is required.

Reports may come from:

Survivors and family members
Readers and researchers
Article subjects and their representatives
Contributors and editors
Journalists, attorneys, archivists, and public officials
Automated security or integrity systems
Moderators who identify a problem during routine review

Anonymous reports are accepted. Contact information can help moderators ask focused follow-up questions, but lack of a name does not make a report invalid. Anonymous claims still require independent verification.

Use the EpsteinWiki contact page for editorial, privacy, or moderation concerns. Use Submit an Epstein Tip for new evidence. Do not send illegal material, passwords, unredacted identity documents, or unnecessary private data through either channel.

What Should Be Flagged

Flag content or conduct involving:

Immediate threats, stalking, intimidation, or incitement
Suspected child sexual abuse material or links that appear to lead to it
The identity of a survivor or minor who has not clearly made that identity public
Home addresses, personal phone numbers, private email addresses, medical data, financial account details, or credentials
A living person confused with someone else
An unsupported allegation of criminal, sexual, financial, or professional misconduct
A quotation, filing, image, or excerpt materially altered or stripped of decisive context
A false description of a charge, verdict, dismissal, settlement, appeal, or court order
Sealed, restricted, privileged, or unlawfully obtained material that presents a publication risk
Fabricated documents, synthetic media, misleading edits, or false provenance
Harassment, slurs, threats, brigading, impersonation, or retaliation
Copyright infringement, plagiarism, or removal of source attribution
Compromised accounts, suspicious access, malicious links, or exposed credentials
Moderator conflicts, hidden reversions, or misuse of access
Broken source links, OCR errors, duplicate pages, formatting problems, and other routine defects

When in doubt, flag the item and describe the uncertainty. Do not increase its reach to ask the crowd whether it is authentic.

What Should Not Be Flagged as Misconduct

Do not use moderation to suppress:

Accurate, relevant, and properly sourced reporting because it is embarrassing
A good-faith editorial disagreement
A documented denial, defense, acquittal, dismissal, or correction
A source merely because it contradicts a preferred theory
Criticism of institutions, public officials, litigation conduct, or published scholarship
Careful language distinguishing allegations from findings
A contributor’s refusal to publish personal information that is not necessary to the story
A request for stronger sourcing or more precise wording

A person may still ask for review of accurate content on privacy, proportionality, or safety grounds. Moderators should consider that request, but EpsteinWiki does not promise removal solely because a documented fact is unwanted.

Severity Levels

Level	Typical concerns	Initial response
Level 1: Critical	Immediate danger, credible threat, suspected child sexual abuse material, exposed survivor or minor identity, active doxxing, leaked credentials, serious system compromise, or clear violation of a sealing restriction	Contain immediately, limit access, notify a senior moderator, preserve only lawful evidence, and direct emergencies to the appropriate authority
Level 2: Urgent	Serious misidentification, unsupported criminal allegation about a living person, sensitive personal data, manipulated evidence, impersonation, targeted harassment, or high-risk legal demand	Temporarily restrict the affected portion when necessary and begin expedited two-person review
Level 3: Priority	Material source failure, incorrect court status, omitted denial, misleading headline or image, plagiarism, coordinated misinformation, or repeated policy breach	Place in the priority queue, add a temporary caution if readers could be misled, and assign an uninvolved reviewer
Level 4: Routine	Broken link, minor date error, typo, formatting defect, duplicate entry, bad tag, or non-material OCR error	Correct through ordinary editing and log the change when it affects meaning

Severity is based on likely harm, not the prominence of the complainant. A survivor’s exposed identity may require faster action than a lawyer’s broad demand concerning a public figure.

Level 1: Immediate Safety Response

If a report indicates immediate danger to any person, site moderation is not an emergency service. In the United States, call 911 or local law enforcement. Outside the United States, contact the appropriate local emergency service.

For a threat connected to a possible federal crime, the FBI advises contacting a local field office or submitting information through tips.fbi.gov. The FBI also advises preserving electronic threats rather than deleting them.

The moderator should:

Avoid engaging or provoking the threatening person.
Restrict public exposure to private addresses, schedules, contact details, or other targeting information.
Preserve the exact message, URL, username, date, time, and available headers when lawful and safe.
Notify the person at risk when doing so will not increase danger.
Escalate internally to the senior safety contact.
Record what was shared with an outside authority, by whom, when, and under what basis.

Do not promise confidentiality that the site cannot legally or technically guarantee.

Suspected Child Sexual Abuse Material

If material appears to depict the sexual exploitation of a child, do not download it, duplicate it, forward it to colleagues, upload it to a moderation folder, or attempt amateur authentication. Stop viewing it and restrict access through the safest available administrative control.

Record only the minimum information needed to identify the location, such as the URL, account, file identifier, time discovered, and how it appeared. Do not place an explicit description or thumbnail in an ordinary moderation ticket.

In the United States, suspected online child sexual exploitation may be reported to the National Center for Missing & Exploited Children’s CyberTipline. NCMEC also operates the 24-hour number 1-800-THE-LOST. If a child is in immediate danger, contact emergency services.

Moderators must not investigate the depicted person’s age by circulating the material. Referral, containment, and non-distribution are the priorities.

Survivor Identity and Privacy

The public availability of a name somewhere on the internet does not automatically authorize EpsteinWiki to amplify it. Before identifying a survivor, editors must consider whether the person has self-identified publicly, whether the name is necessary to understand the record, whether a court has protected the identity, and whether publication could expose the person to retaliation or renewed harassment.

Flag immediately when content:

Connects a pseudonym to a legal name without a clear editorial basis
Reveals the identity of a person who was a minor at the time of abuse
Combines scattered facts to defeat an existing redaction
Publishes a survivor’s address, workplace, family details, health information, or current location
Uses a survivor’s image without context or permission where the use creates foreseeable harm
Treats testimony as entertainment or invites readers to judge trauma responses

Follow the Trauma-Informed Writing Toolkit, Working With Survivor Testimony, and Privacy Safeguards for Minors.

Doxxing and Personal Information

Doxxing includes publishing or assembling information to enable intimidation, unwanted contact, surveillance, fraud, or physical targeting. The information may be accurate and still be unsafe.

Remove or restrict, unless clearly necessary and lawfully publishable:

Residential addresses and real-time locations
Personal telephone numbers and private email addresses
Social Security, passport, driver’s license, or tax identifiers
Bank, card, routing, and login information
Passwords, recovery codes, tokens, and security answers
Medical details unrelated to a documented public-interest issue
Names, schools, or locations of minors
Family information used as leverage against a contributor, source, or survivor

Public business addresses, court-listed service addresses, and official office numbers may sometimes be relevant. Editors should still ask whether repeating them adds evidence or merely increases exposure.

Misidentification

Misidentification is among the fastest ways an evidence archive can harm an innocent person. Similar names, initials, OCR mistakes, reused email aliases, family relationships, and incomplete address-book entries can all create false matches.

An urgent review should compare:

Full legal name and known variants
Age and relevant dates
Employer, title, and location at the time
Contact details shown in the source
Photograph provenance
The document’s own context
Independent records that confirm or contradict the match

Do not treat matching names as matching identities. If identity remains uncertain, say so clearly or remove the identification. A correction should reach the headline, tags, captions, metadata, redirects, social previews, and any network map that inherited the error.

Allegations About Living People

Articles concerning living people require exact attribution. The source must support the precise claim, not a nearby or more dramatic version of it.

Moderators should flag language that converts:

Appearance in a record into participation in a crime
An allegation into an established fact
A civil complaint into a criminal charge
A charge into a conviction
A settlement into an admission of liability
Social contact into knowledge of abuse
Travel on one route into presence at another place
A lawyer’s argument into a judicial finding
A disputed statement into a consensus

Where a serious allegation is included, the article should identify who made it, where it appears, its procedural status, any relevant denial, and what has or has not been adjudicated.

Association Is Not Culpability

Epstein-related records contain thousands of names. A person may appear because of a telephone call, invitation, business contact, address-book entry, staff task, legal role, social event, academic meeting, financial transaction, or third-party mention.

Moderators must intervene when an article uses proximity as a substitute for proof. The sentence should state the documented connection and stop at the edge of the evidence.

Good: “The calendar lists a meeting with X on May 4, 2015. The record does not state who attended or what was discussed.”

Bad: “X was part of Epstein’s trafficking network,” when the only source is a calendar entry.

This distinction protects innocent people and makes evidence of genuine facilitation more credible.

Threats, Harassment, and Retaliation

EpsteinWiki does not tolerate threats, stalking, sexual harassment, discriminatory abuse, coordinated targeting, impersonation, blackmail, malicious exposure of personal information, or retaliation for a good-faith flag.

Moderators should separate criticism from harassment. Strong criticism of an edit is allowed. Repeatedly contacting a contributor after being told to stop, threatening employment or family, publishing private details, or directing a crowd toward an individual is not.

Preserve the full context of a threat, including surrounding messages and timestamps. Do not require the target to negotiate with the sender. Apply the Harassment & Threat Response for Contributors and Digital Safety for EpsteinWiki Contributors & Researchers.

Account restrictions may be imposed before the underlying editorial dispute is resolved when conduct itself presents a safety risk.

Manipulated Documents, Images, Audio, and Video

Flag media when there is reason to believe it was fabricated, selectively edited, mislabeled, generated, altered, or presented with false provenance.

Reviewers should seek:

The earliest available version
The original hosting or filing source
Complete pages before and after an excerpt
File metadata, while recognizing metadata can be changed
Consistent pagination, fonts, stamps, signatures, and exhibit labels
Independent copies or references
A documented chain from source to publication
Expert analysis when the issue exceeds editorial competence

Do not call a document fake merely because it is inconvenient, or authentic merely because it looks official. If authenticity remains unresolved, label the uncertainty and prevent the item from carrying more weight than the evidence permits. Consult the Disinfo & Manipulated Media Guide and Document Redaction Integrity.

Redactions and Reconstruction Risk

A redaction is not an invitation to solve for the hidden text. Different documents can reveal fragments that, when combined, identify a protected person. That reconstruction may defeat a court order, survivor safeguard, or responsible editorial choice.

Flag:

Failed digital redactions where hidden text remains selectable or recoverable
Filenames or metadata that reveal a protected identity
Captions that identify a person obscured in an image
Side-by-side records assembled to expose a pseudonym
OCR that restores text intentionally removed from the public copy
Cropping that conceals the fact that a source was redacted

Moderators should preserve the public record’s meaning without reverse-engineering protected information.

Court Orders, Sealed Records, and Restricted Material

Legal markings require careful reading. “Under seal,” “confidential,” “attorneys’ eyes only,” and “protective order” can have different effects depending on the court, the person bound by the order, later unsealing, and the source of the copy.

If a flag alleges a sealing or protective-order problem:

Restrict the disputed file when credible harm could occur.
Locate the operative order and docket history.
Determine whether the exact material was later unsealed or released by an authorized source.
Identify whether EpsteinWiki or the submitter is bound by an order.
Seek qualified legal review when the answer is unclear.

Do not accept a bare assertion that a public court filing is “confidential.” Do not ignore a verified restriction because the file has already circulated elsewhere.

Copyright, Plagiarism, and Attribution

EpsteinWiki may quote, summarize, analyze, and link to source material, but contributors must not copy entire copyrighted works merely because they are relevant. Court filings and government records may have different copyright status from journalism, books, photographs, databases, or documentary footage.

Flag:

Large unattributed passages
Images or scans without source and rights information
Removed watermarks or credits
A full paywalled article reproduced instead of summarized
False claims that a contributor created another person’s research
Repeated copyright complaints involving the same source or account

A copyright complaint should identify the protected work, the challenged location, the claimant’s relationship to the work, and the requested action. Moderators should preserve citations and factual analysis while removing or replacing material that cannot be responsibly republished.

Security Incidents and Compromised Accounts

Treat unexpected password changes, unknown administrative edits, malicious redirects, exposed tokens, suspicious bulk downloads, phishing messages, or impossible login patterns as security flags.

Immediate actions may include:

Revoke active sessions and exposed credentials
Temporarily lock the affected account
Preserve access logs
Remove malicious links or code
Verify the account owner through a trusted channel
Review recent edits and exports
Require a new unique password and multi-factor authentication
Notify people whose private data may have been exposed

Never paste a password, private key, session token, or recovery code into a moderation ticket. Record that a credential was exposed, not the credential itself.

Reporting a Flag

A useful report is specific, restrained, and reproducible. Include:

Exact page URL or file identifier
Section heading, paragraph, caption, timestamp, or edit revision
Category of concern
The exact words or conduct at issue
Why it may violate policy
Links to stronger or contradictory sources
Whether anyone faces immediate harm
Whether personal information is exposed
Your preferred contact method, if you want a response

Do not attach unnecessary copies of sensitive documents. Do not forward threatening or illegal files to multiple recipients. A concise report with an exact location is more useful than a broad accusation against the site or contributor.

Example: “On the X article, under ‘2018 Meeting,’ the second paragraph says Y was convicted. The linked docket shows the charge was dismissed on June 4, 2020. Please correct the procedural status.”

Intake and Acknowledgment

The receiving moderator creates one case record and assigns a unique identifier. Duplicate reports should be linked rather than treated as independent proof.

The intake record should capture:

Date and time received
Reporter contact, if supplied
Affected content and revision
Initial category and severity
Immediate safety concerns
Temporary action taken
Assigned reviewer
Related prior reports

When practical, acknowledge receipt without promising an outcome or exact completion time. Acknowledgment should not repeat sensitive content. If the report is misdirected, route it internally rather than requiring a survivor or threatened contributor to tell the entire story again.

Triage

The first reviewer should perform a narrow triage, not a full trial of every issue raised.

Ask:

Is there immediate physical danger?
Does the site currently expose illegal material, a protected identity, credentials, or actionable personal data?
Could continued publication cause severe and difficult-to-reverse harm?
Is the claim about a living person materially unsupported?
Is there evidence of account compromise or coordinated harassment?
What is the smallest temporary action that contains the risk?

Reclassify the case if new facts change the risk. A routine broken link may become urgent if it redirects readers to malicious content.

Temporary Protective Actions

Moderators may take temporary action before reaching a final decision. Available measures include:

Hide a sentence, image, attachment, or revision
Remove a page from search and social previews
Replace a download with a review notice
Revert to the last stable version
Disable comments or editing on a targeted page
Restrict an account or require reauthentication
Add a visible accuracy or authenticity warning
Limit internal access to a high-risk file

The action must be no broader than reasonably necessary. Record who authorized it, why, what was affected, and when it will be reviewed. Temporary restriction does not prove wrongdoing and should not be described publicly as a final finding.

Preserving Evidence Safely

Moderation evidence should allow a later reviewer to understand what happened without needlessly retaining harmful material.

For lawful content, preserve:

The URL and revision identifier
Date and time, including time zone
Exact text at issue
Relevant surrounding context
Source links and docket references
A screenshot when necessary
File hash and metadata when authenticity is disputed
Edit and access logs available through site systems
Communications about the decision

Minimize personal data and restrict access according to need. Do not create extra copies of illegal or exploitative material. Follow Evidence Handling 101 and Handling Sensitive Material.

Assigning the Review

Routine issues may be resolved by one editor. High-risk cases require at least two qualified reviewers when practical, including one senior editor for:

Survivor identification
Serious allegations about living people
Threats and doxxing
Account termination
Authenticity findings about consequential evidence
Legal demands or alleged court-order violations
Reversal of a prior high-risk decision

The reviewer should have access only to the information needed for the task. Technical, legal, trauma-informed, or subject-matter expertise may be requested. Expertise informs the editorial decision but does not replace the written evidence analysis.

Conflicts of Interest and Recusal

A moderator must disclose and step back from a case when they:

Wrote or substantially edited the disputed content
Submitted the flag
Have a personal, financial, political, professional, or legal relationship with a party
Have publicly campaigned for a specific outcome in the same dispute
Are a witness or source in the underlying matter
Have been threatened by, or are in an active dispute with, the affected contributor

Mere familiarity with the subject is not always disqualifying, but the relationship must be disclosed. The case log should name the replacement reviewer. Urgent containment may be performed despite a conflict when no one else is available, but the conflicted moderator must limit action to safety and transfer the decision promptly.

The Review Process

The assigned reviewer should:

Define the exact disputed claim or conduct.
Retrieve the challenged revision and its sources.
Read the complete relevant source, not only the quoted fragment.
Check dates, names, jurisdiction, docket posture, and document provenance.
Separate verified fact, allegation, testimony, inference, and unknown.
Locate contradictory evidence and any relevant response or denial.
Assess privacy, safety, legality, and public interest.
Ask the contributor focused questions when appropriate.
Obtain a second review when required.
Select the least restrictive outcome that resolves the problem.
Record reasons and required follow-up.

When sources conflict, do not choose the most dramatic. Apply Handling Contradictory Evidence and explain the conflict in the article when it matters.

Contacting Contributors and Article Subjects

Contributors should normally be told which passage is under review and invited to supply the source or reasoning behind it. A deadline may be set according to urgency. Lack of response does not prove misconduct, but moderators may act on the available record.

For serious, newly published allegations about a living person, editors should consider a request for comment before publication or during review. The request must state the substance accurately, allow a reasonable chance to respond, and avoid disclosing survivor information or confidential sources.

Do not contact a suspected harasser, abuser, or threatening person when doing so could endanger someone. Do not reveal who filed a confidential report unless necessary, lawful, and authorized.

Decision Standards

A moderation finding should use the applicable standard, not intuition alone.

Accuracy: Does the cited evidence support the exact words used?

Attribution: Can a reader tell who alleged, testified, argued, found, denied, or verified the claim?

Provenance: Can the document or media be traced to a credible origin?

Context: Would omitted information materially change a reasonable reader’s understanding?

Privacy: Is the personal detail necessary and proportionate to the public interest?

Safety: Could publication foreseeably facilitate harm?

Legality: Is there a credible legal restriction or prohibited material issue?

Conduct: Did an account engage in harassment, deception, retaliation, or repeated disregard of policy?

Uncertainty should be written as uncertainty. Moderators are not courts and should not invent findings beyond the record.

Available Outcomes

A review may result in:

No action: The content complies with policy.
Technical repair: Fix a link, OCR error, tag, date, or formatting defect.
Correction: Replace a factual error and record the change.
Context added: Supply procedural status, surrounding text, denial, limitation, or explanation.
Attribution strengthened: Identify the speaker, filing, court, or evidentiary status.
Warning label: Mark disputed authenticity, graphic content, or unresolved provenance.
Redaction: Remove personal or protected information while retaining the relevant record.
Restriction: Limit access, indexing, download, commenting, or editing.
Removal: Delete material that cannot be safely or responsibly published.
Merge or redirect: Consolidate duplicative or misleading pages.
Account action: Warn, limit, suspend, or ban an account.
External referral: Direct an emergency, exploitation report, threat, or cybercrime report to the proper authority.

The written decision should connect the outcome to the evidence and rule.

Corrections, Notes, and Version History

Material corrections should be visible enough to reach readers who encountered the error. Quiet edits are appropriate for spelling, formatting, and other changes that do not alter meaning.

A correction or editor’s note should state:

What was wrong or incomplete
What changed
When it changed
The source supporting the correction
Whether the error affected a headline, image, identity, or central conclusion

Do not repeat defamatory or private information in the correction note. Preserve internal revision history even when a dangerous public revision must be suppressed. Follow the Corrections Policy and the site’s version-control procedures.

Account Warnings, Restrictions, and Bans

Account action is based on conduct and risk, not viewpoint. Factors include:

Severity and immediacy of harm
Whether the action was deliberate, reckless, or accidental
Prior warnings and repeated patterns
Attempts to conceal edits or evade restrictions
Retaliation against reporters or moderators
Willingness to correct the problem
Risk to survivors, minors, sources, contributors, and site systems

A warning should identify the rule and expected correction. Suspension may be appropriate for repeated violations or a serious incident requiring investigation. Permanent bans are reserved for severe or persistent abuse, credible threats, deliberate exposure of protected people, distribution of prohibited material, account compromise activity, or systematic bad-faith manipulation.

Ban evasion may be treated as a separate violation.

Notification of the Decision

When practical and safe, notify the flagger and affected contributor that review is complete. The notice should summarize the issue, outcome, and any public correction or appeal path.

The notice may withhold:

Survivor or minor identities
Private contact information
Security methods and access logs
Confidential source details
Information whose disclosure could enable harassment
Privileged legal advice
Details restricted by law or court order

Moderators should not disclose a reporter’s identity merely to satisfy curiosity. A person accused of misconduct should receive enough information to understand and answer the allegation unless disclosure would create a safety or privacy risk.

Appeals and Reconsideration

A contributor, flagger, or directly affected person may request reconsideration when they identify:

A factual error in the decision
Material evidence not previously reviewed
A conflict of interest
Misapplication of policy
A disproportionate remedy
Changed circumstances, such as an unsealed record or corrected source

An appeal should name the case or page, state the challenged outcome, and provide the specific reason. It should be reviewed by someone who did not make the original decision whenever practical.

Appeal is not an opportunity to repeat abuse, contact a survivor, or mobilize harassment. Temporary safety measures remain in effect while an appeal is pending unless the appeal reviewer changes them.

Repeat, Coordinated, and Abusive Flagging

Good-faith reports are protected even when moderators find no violation. A mistaken flag is not abuse.

Abusive flagging may include:

Repeatedly reporting the same resolved issue without new evidence
Flooding the queue to suppress documented reporting
Coordinating mass reports based on false instructions
Using reports to expose private information
Threatening moderators unless content is removed
Impersonating a survivor, lawyer, rights holder, or public official
Filing knowingly false emergency claims

Moderators should consolidate duplicates, preserve evidence of coordination, and continue to evaluate the underlying content independently. A bad-faith campaign does not make the challenged article accurate. Review the evidence anyway.

Anti-Retaliation Rule

No one may retaliate against a person for making a good-faith report, participating in review, correcting an article, preserving evidence, declining an unsafe assignment, or appealing a decision.

Retaliation includes demotion, exclusion, threats, doxxing, hostile contact, smear campaigns, selective enforcement, exposure of confidential identity, or manipulation of account permissions.

Report retaliation as a new high-priority case. Assign it to a moderator outside the original conflict. When the alleged retaliator has administrative access, preserve logs and restrict that access as necessary while review proceeds.

Confidentiality and Data Minimization

Moderation records often contain more sensitive information than published pages. Access must be limited to people with a work-related need.

Collect only what is necessary. Avoid copying full medical records, identity documents, financial data, private family communications, or graphic descriptions into case notes. Use pseudonyms or case identifiers where possible.

Retention should reflect legal obligations, security needs, appeal periods, and the risk of retaining the material. When a case no longer requires sensitive attachments, securely remove them while preserving a non-sensitive decision record. Never use moderation files for curiosity, personal research, or unrelated publication.

Moderator Records and Audit Trail

Each significant case should record:

Case identifier and dates
Reporter and affected account, using protected identifiers when needed
Page, file, revision, or conduct reviewed
Severity and category changes
Temporary actions
Reviewers and recusals
Sources examined
Findings and unresolved questions
Final outcome and policy basis
Notifications sent
Appeal and follow-up status

Administrative actions should be attributable to an individual account. Shared moderator logins are prohibited. Logs must not be altered to hide a mistake. Correct a mistaken record with a dated amendment.

Transparency Without Exposure

EpsteinWiki should periodically publish aggregate information about moderation when feasible, such as the number of flags received, broad categories, corrections issued, content restrictions, account actions, and appeals.

Transparency reporting must not reveal:

Survivor or minor identities
Information that reconstructs a redaction
Security weaknesses
Confidential sources
Personal data about reporters or contributors
Details likely to intensify an active threat

Aggregate reporting helps readers see whether the system works without turning private harm into public spectacle.

Moderator Wellbeing

Moderators may encounter threats, graphic descriptions, survivor testimony, and manipulative communications. No moderator is required to view potentially illegal imagery or repeatedly handle traumatic content.

Teams should use rotation, limited exposure, protected breaks, debriefing, and permission to recuse for wellbeing. Graphic files should not autoplay or appear as ticket thumbnails. Warnings should describe the category without sensational detail.

A distressed moderator should transfer the case, document what remains, and seek support. Continuity matters, but no case depends on one person absorbing unlimited harm.

Decision Matrix

If the review finds	Preferred action	Additional requirement
Accurate, relevant, and safely presented content	No action	Record why the flag was closed
Minor non-material defect	Repair	Note in ordinary edit history
Material factual error	Correct promptly	Add a correction note when readers may have relied on it
Supported claim with missing context	Add context or attribution	Review headline, caption, and metadata too
Exposed private or protected information	Redact or restrict	Check cached previews, attachments, and linked pages
Serious claim unsupported by cited evidence	Remove or rewrite	Do not leave the accusation in tags or summaries
Authenticity unresolved	Label and limit evidentiary weight	Describe what is known and unknown
Manipulated or fabricated material	Remove or clearly debunk	Preserve lawful forensic record and trace dependent pages
Threat, doxxing, or targeted harassment	Protect the target and restrict conduct	Preserve evidence and consider outside reporting
Account compromise	Secure account and review activity	Revoke sessions and inspect affected edits
Repeated deliberate violations	Suspend or ban	Provide appeal route unless safety prevents contact

Moderator Checklist

Before closing a significant flag, confirm:

The exact content and revision were identified
Immediate safety and illegal-material risks were assessed
Temporary measures were documented and revisited
The strongest available sources were read in context
Identity was verified rather than inferred from a name match
Allegation, testimony, charge, finding, and conviction were not conflated
Relevant denial or contradictory evidence was considered
Survivor and minor privacy was protected
A conflicted reviewer recused
A second reviewer handled high-risk issues
The remedy was no broader than necessary
Headline, tags, captions, metadata, and related pages were checked
The contributor and flagger were notified when appropriate
Correction, decision, and appeal records were completed

Frequently Asked Questions

Does a flag remove an article automatically?

No. It opens review. Moderators may temporarily restrict a specific portion when continued publication creates serious risk.

Can an article subject demand removal?

Anyone may request review. The subject’s status does not control the outcome. Editors assess accuracy, sourcing, privacy, safety, law, and public interest.

Can anonymous reports be trusted?

They can contain valuable information, but moderators verify the evidence independently. Identity is not a substitute for proof.

What if the source is a court filing?

A filing proves that a statement was filed. It does not automatically prove every allegation in it. Identify the speaker and procedural status.

What if information is already public?

Public availability matters, but it is not the only consideration. Amplification can create new harm, especially for survivors, minors, home addresses, and reconstructed redactions.

What if moderators disagree?

Record the disagreement, maintain necessary temporary safeguards, and seek an uninvolved senior or specialist review. Do not resolve uncertainty through an edit war.

How fast will EpsteinWiki respond?

Cases are prioritized by risk. Critical safety and privacy issues require immediate containment. Routine editorial reports enter the normal queue. EpsteinWiki should acknowledge reports when practical but cannot promise a fixed outcome time for every investigation.

Related EpsteinWiki Policies and Guides

External Safety and Reporting Resources

These links are provided for public information. EpsteinWiki does not control outside agencies, cannot guarantee an investigation, and is not a substitute for emergency services, legal advice, medical care, or crisis support.

Closing Principle

Moderation is not the opposite of transparency. Done correctly, it is what makes transparency durable.

The archive becomes more credible when it can correct itself, more useful when readers can distinguish proof from proximity, and more humane when survivors do not have to surrender their privacy to enter the historical record. A flag should never become a shortcut to censorship, and publication should never become an excuse to ignore foreseeable harm.

The standard is demanding because the subject demands it: preserve the evidence, protect people, show the reasoning, and publish only what the record can carry.

Editorial Note

This is a living policy. It should be reviewed after major incidents, material legal changes, security events, or recurring moderation failures. Substantive revisions should be dated and summarized so contributors can understand what changed.

Suggested review cycle: At least annually, and after any Level 1 incident.

Policy owner: EpsteinWiki editorial and moderation leadership.

Applies to: Contributors, editors, moderators, administrators, contractors, and official EpsteinWiki channels.

PostedDecember 6, 2025